|
Misogynist posted:It seems obvious to me that this is targeted at netbooks and similar form-factors, and I don't know why people are putting so much effort into dissecting how horrible it is at replacing their GPGPUs and video encoders. Probably because the thread title proclaims it will "change the face of computing". B-Nasty posted:Personally, I see a future where computers are based around 2 separate main chips. One that is blazingly fast at serial, or not quite parallel tasks, and one that can handle those (comparably) rare scenarios that benefit from massive, fully-independent parallelization. You're sort of wrong. It's only going to be a single chip, and we can already see it's first stage in the form of Turbo Boost. The future is a single chip which can reconfigure itself for better parallel or serial throughput based on the current task.
|
#
¿
Sep 15, 2010 20:20
|
|
|
|
| # ¿ Apr 24, 2024 11:37 |
|
|
fishmech posted:or did you really think binning only ever happened to defective chips? Binning happens to all chips (although for that matter, all chips are defective to some extent). What you're complaining about is selling chips at a lower rating than the bin they qualified for.
|
|
#
¿
Sep 19, 2010 23:05
|
|
|
Spime Wrangler posted:The i5 is an amazing processor, buy one now and it'll last you plenty of time. i5 is a branding, not a processor. Many of the Sandy Bridge processors will be sold as i5. And they'll share the two features that make the i5-750 so compelling- quad cores with turbo boost. Spime Wrangler posted:Also, those benchmarks that were linked show sandy bridge underperforming the top-end i7 chips. The benchmarked sandy bridge is targeting one step below the i5-750, so about $175. The high-end i7 that it can't quite beat is $1,000 and holds a 2 core advantage. Sandy Bridge probably won't become a must buy until they come out with an affordable 6-core turbo part, but it's still looking pretty good.
|
|
#
¿
Sep 20, 2010 00:26
|
|
|
MachinTrucChose posted:New architecture with 10% power savings and 10% speed improvement That's 10% speed improvement per cycle, but you'll get more cycles at a lower price; so more like 25-30% better at the same price point (depending on how well the new turbo boost works out). Spime Wrangler posted:So I'm not a computer engineer or anything (so I should probably shut up) but given the level of architectural rearrangement we're seeing with CPUs today I think it's somewhat natural that the sockets obsolete relatively quickly. You're actually right, although it's not the GPU that's driving it. The last sockets truly bridging several architectures were LGA 775 on the Intel side, and Socket A on the AMD side. Coincidentally, those worked with the last architectures before they moved to on-die memory controllers. On the AMD side now, they've been working to maintain some compatibility in spite of this, which you can see from the AM2/AM2+/AM3 sockets. Intel hasn't bothered with it, which is why we get the 1156/1155 forced incompatibility.
|
|
#
¿
Sep 21, 2010 06:18
|
|
|
I bought a "huge" Thermalright AX-7 for my Athlon... it was big enough for an 80mm fan! I had to dremel off the bottom of a couple fins to fit in on my motherboard.
|
|
#
¿
Sep 23, 2010 04:32
|
|
|
If you want to take apparent slowness out of a computer, just get an SSD.
|
|
#
¿
Sep 23, 2010 14:57
|
|
|
wolrah posted:There's rumors that the next version of Windows will require UEFI, though I doubt that will be the case since almost no one has it right now so very few computers could be upgraded. It would be nice if they'd come out and say that the version after will require it though, that gives plenty of time for hardware manufacturers to get up to speed and for most computers that would be worth upgrading the OS on to have support. The only thing companies will do if you give them a couple extra years like that is spend a couple years hoping you change your mind.
|
|
#
¿
Oct 9, 2010 00:39
|
|
|
movax posted:These are going to run cooler than their Conroe predecessors anyways, correct? No, not at all. In fact, if you're moving from an E6600 to an i5-2500, you'll be moving up from the 65W TDP bracket to the 95W TDP bracket (since you're adding two cores). CPUs are TDP limited these days, so even across architectures their TDP targets are the same.
|
|
#
¿
Dec 14, 2010 23:19
|
|
|
freeforumuser posted:I'm sure retail SBs definitely won't hit the overclocks achieved by the hand-picked ES chips. It sounds like you're under the impression that engineering samples are of higher quality than the retail chips. They aren't. Tweaks to the design to improve yields will have been added, and the manufacturing process will have been improved, so a substantially larger portion of the retail chips will be capable of achieving those speeds than the engineering samples, even at release. I would expect Intel to favor higher-binning chips for the unlocked models anyway, since it wouldn't cost them anything and better overclocks will convince more people to pay the extra.
|
|
#
¿
Dec 28, 2010 20:33
|
|
|
Combat Pretzel posted:All upcoming SB board from Asus do have EFI. And yes, you need EFI for +2TB drives, at least if you don't want to partition it. If the boot partition's below 2TB, then it should be OK I think. If you want to partition more than 2TB of space, even on a single partion, you have to use GPT. You cannot boot to a GPT partition with a BIOS, so you can't use more than 2TB on your boot drive with BIOS. mobby_6kl posted:but everyone is also forgetting about the page file. I have 6 gigs of RAM installed now, and according to windows 6143 MB are allocated right now, with 9214 being recommended. Windows' recommendations there are kind of out of date. It depends on what you're running, but 1-2GB is probably a better size for that much RAM.
|
|
#
¿
Dec 30, 2010 03:11
|
|
|
Nonpython posted:
The only difference between Linux and Windows here is that Windows guarantees adequate disk space to successfully hibernate, whereas Linux does not (necessarily).
|
|
#
¿
Dec 30, 2010 03:56
|
|
|
And an extra 12GB of disk cache will go a long way to hide the misery and suffering of mechanical hard drives if you haven't moved to an SSD yet.
|
|
#
¿
Jan 7, 2011 16:19
|
|
|
Inept posted:No, don't tell people to do that. That is some stupid "optimization" advice that was touted a decade ago and does not apply. Having a page file available won't make anything slower. And it was wrong a decade ago, and has become increasingly wrong since then. Not having a page file will make things slower, even if you have ridiculous amounts of RAM and nothing actually needs to be paged out.
|
|
#
¿
Jan 7, 2011 20:27
|
|
|
movax posted:The IGP hardware accelerated decoding (IIRC) will put out 24.000, not 23.976. It's such a loving It's in the IGP output, not the decoding. It doesn't matter whether or not you're using software decoding, it's just plain incapable of sending the right signal to the display. But apparently they have a software patch that will let you do 23.97hz, which comes out to dropping a single frame about every 2 1/2 minutes, so I'd still call it a pretty loving  complaint. Plus the low-end HTPC targeted nvidia and ATI cards aren't exactly costly if it really bothers you that much.
|
|
#
¿
Jan 31, 2011 22:23
|
|
|
It could also be an issue with the fan/fan speed control.
|
|
#
¿
Feb 20, 2011 22:08
|
|
|
Combat Pretzel posted:Intel SATA 3Gb ports vs the Marvell 6Gb. Which ones are the better option? With 3Gb drives? I keep noticing that Marvell gets a small bad rap all the time. Why's that? Aside from the poor performance already mentioned, they don't have particularly stellar reliability. I had stability problems on my last system which I eventually determined coincided perfectly with having anything (even a DVD drive) on the marvell controller's ports. A number of people have posted in the SSD thread with similar issues.
|
|
#
¿
Apr 8, 2011 14:45
|
|
|
Verizian posted:Crysis 2 is 32bit so it can't see more than 3.5GB including video ram. And relatively few games are large address aware, in which case they're stuck with 2GB. Verizian posted:That said 8GB is nice for doing other things at the same time as gaming as long as they are not HDD bound. Maybe you mean CPU bound? HDD bound things will benefit the most from more RAM, since more memory means less paging, and a larger disk cache.
|
|
#
¿
Apr 12, 2011 17:39
|
|
|
Henrik Zetterberg posted:It mostly depends on where physically the core is located. The core right next to GT should theoretically be the warmest since it has more "stuff" going on near it. Or if you're not using GT, the middle two will probably be the warmest. Also, the thermal sensors aren't exactly carefully calibrated scientific instruments. Even if the cores are all exactly the temperature they still aren't going to read out the same.
|
|
#
¿
Apr 16, 2011 00:55
|
|
|
DaNzA posted:Guessing that they might not have the chip nailed down yet and don't want to repeat the Pentium III 1.13Ghz thing again. There's little risk of ever having that happen again. Chips are now limited by TDP, and can easily exceed the rated speeds using more power (thus why they easily overclock so far). They may be binning the chips lower because not enough of them can meet the target TDP at 4GHz.
|
|
#
¿
May 7, 2011 23:54
|
|
|
Agreed posted:So... Help me out, it's A Bad Thing but overclocking in general is bolting big gently caress-off heat sinks and taking extraordinary measures to rip stability and performance out of exceeding the factory specifications, so isn't overclocking in general, if it exceeds stock voltage, A Bad Thing? For starters, all modern processors (pretty much everything made since the Pentium 4 or Athlon 64) is TDP limited. In other words, the speed of the processor is limited not by the capabilities of the chip, but by an agreement not to dissipate more than a certain amount of energy. An overclocker with a big heatsink can essentially tell the processor it's okay to violate that agreement, and achieve a substantial speed increase without exceeding the electrical specifications of any of the components. After that, there are different levels of badness. When you first start exceeding the specifications, you're just eating into safety margins, put there to minimize the change of marginal chips failing under poor conditions (such as motherboards that do poor power conditioning, like LLC). Past that, the increased current accelerates electromigration, reducing the lifespan of the chip. Since the rated lifespan generally greatly exceeds the useful lifespan, this generally isn't that big of a deal. If you push the voltage even higher, then you start hitting what Alereon was calling a Bad Thing(tm). Sudden voltage changes or excessively high voltages can cause transistors to open or close inappropriately, causing instability, or they can cause electricity to travel across barriers it's not intended to travel (generally destroying said barrier in the process), or excessively high current resulting from the voltage can cause various kinds of thermal runaway, destroying transistors or wires.
|
|
#
¿
Jul 26, 2011 21:30
|
|
|
Straker posted:People used to be scared of using heatsinks so big to begin with, too, that only explains away part of it but probably didn't help adoption. It wasn't just a matter of fear, it was also the mechanical specifications of the sockets. The sockets and motherboards weren't designed to support the force necessary to get good mounting pressure with large heatsinks, and they only required enough clearance to fit 60mm heatsinks. Modern socket specifications give heatsink manufacturers much more room and much better mounts than in the past.
|
|
#
¿
Aug 7, 2011 05:12
|
|
|
To be fair, even if the chip supports a feature, there is a non-zero cost to validating the feature (and rejecting failures) and offering support for the feature. Reading about the power savings and GPU performance makes me feel bad about just buying a sandy bridge laptop. Curse you and your continual advancement, technology!
|
|
#
¿
Sep 18, 2011 00:10
|
|
|
Shaocaholica posted:So have all previous Intel 6 core procs actually been 8 core procs with 2 turned off? No, this is their first such consumer processor. A few of the recent 6-core Xeons have been die-harvested 8 or 10 core processors, although there have been true 6 core Xeons as well. Shipping a part with components disabled without shipping the full version is a fairly unusual; that extra wasted silicon costs a lot of money.
|
|
#
¿
Dec 1, 2011 23:47
|
|
|
Install Gentoo posted:The only CPUs guaranteed to not be rebinned higher level procs with some manner of defect, are the most expensive CPUs in any given family. And even most of the most expensive ones will have some level of defect separating them from the theoretically ideal. Agreed posted:I KNOW, that's the trouble, we have no idea if it's a real thing or if they're just "let's add an sku gently caress it make money money." And I'm the rear end in a top hat that'd pay for the 2700K because I'd be thinking, poo poo, the 2600K is kind of second tier, what if they're bad 2700s? There has always been a mix of 2600Ks that bin at 2700K performance and ones that don't. As they've made process improvements and new steppings to improve yields, the portion that bin at 2700K performance got large enough to reliably meet demand for a 2700K product, so they added the 2700K, and now the best 2600Ks are 2700Ks. But the thing is, thanks to those same improvements, the remaining 2600Ks still end up being better on average than the older 2600Ks anyway.
|
|
#
¿
Dec 3, 2011 00:14
|
|
|
HalloKitty posted:For all you know, they could be binned identically. I mean really, has anyone ever had a problem overclocking their 2600K by 100MHz? Wedesdo posted:I'm very happy with my 4.8 GHz on my 2500k, on air. If they are binning, it's not too much. They criteria they're binning to are an awful lot stricter than your criteria. For starters, they need to give a safety margin to keep them working in much crappier conditions than you're putting them in. But, more importantly, these days processors aren't frequency limited, they're TDP limited. If your 2600K didn't bin high enough to be a 2700K, it's not because it can't run at 2700K speeds, it's because it can't run at 2700K speeds without exceeding the maximum TDP; it might put out 97W instead of the specified 95W. It can go a lot faster if you tell it to, it's just going to take a lot more juice. Shaocaholica posted:How can you tell which procs (SB and IB) are die harvested and which aren't? Are all IB dual core procs moving forward doing to be die harvested quads? Generally speaking, fulfilling your low end parts through die harvesting is not desirable. You sell a lot more low end parts than high end part, but you won't end up with more harvested low-end parts than fully functional high-end parts unless you have really terrible yields. If/once your yields aren't terrible, then you're spending more silicon than necessary to make the low-end parts, hurting the profit margin on your highest volume products. That said, there are non-trivial engineering costs associated with designing another chip configuration, so it may not be worth it for low-volume parts anyway.
|
|
#
¿
Dec 3, 2011 01:04
|
|
|
Shaocaholica posted:So are SB procs varying by a lot within the same design compared to previous Intel procs? I doubt anyone who truly knows the answer to this would be allowed to say, but there's no reason to think it's anything other than no. It's simply the nature of semiconductor manufacturing, the way it's been for 50 years. It's more economical to make things smaller than to make them reliably. Basically, if you could spend $5,000 today trying to manufacture 100 processors, would you rather end up with 90 Core 2 Extreme QX9770s and 8-9 lesser Core 2 Quads, or 60 SB i5s and 20 SB i7s? (Hint: an i5-2400 trounces an "Extreme" QX9770) Shaocaholica posted:Are we at the point with SBE where the variations are becoming less and less? We have always been at that point. Intel isn't going to tell you any more detail than that, though.
|
|
#
¿
Dec 3, 2011 05:44
|
|
|
Economy Clown Car posted:we've been using 1920x1080 rez displays for quite a few years now The "we" there is still a very small minority. Only ~7% of computers have 1920x1080 displays, and even fewer have something higher than that. The closest thing to a "new standard" is 1366x768, which is rapidly becoming the most common resolution (>40% of Windows 7 systems are 1366x768). Edit: and why do people worry so much about mobo future proofing? Not once in the past decade have I ever felt I could get decent value from upgrading a CPU or motherboard without also upgrading the other. Zhentar fucked around with this message at 17:37 on Dec 29, 2011 |
|
#
¿
Dec 29, 2011 17:31
|
|
|
Gigabyte has also repeatedly had problems with AHCI support. For example, Anandtech ran into some here. A couple weeks ago someone in the SSD thread had to go through a lot of song and dance to get their boot time to squeak in under a minute.
|
|
#
¿
Jan 17, 2012 22:02
|
|
|
Alereon posted:I think that explanation is bullshit because DisplayPort and HDMI push video at about the same bitrate through a passive cable with no problem. No, they don't. DisplayPort 1.2, which barely anything can support yet, only pushes 4.32Gb/s, and HDMI 1.3/1.4 can only support a lethargic 3.4Gb/s. They achieve better throughput by having 3-4 lanes at that bitrate.
|
|
#
¿
Jan 19, 2012 16:55
|
|
|
SRQ posted:If I understand this right, they took the original pentium design and die shrunk it + added new instruction sets? Assuming you're talking about Claremont/NTV... what's special is they've managed to dramatically improve the lower bound on idle efficiency. With Sandy Bridge, built on the same process, essentially no matter how low you set the clock speed, you can't drop the voltage below .8v (don't quote me on that exact number). If your processor leaks power at, say, 1 watt at .8v, then the only way to get it to draw less than 1 watt is to turn some or all of it off entirely. Turning part or all of the processor on or off wastes some power in the process, and carries a small performance penalty (I think the major transitions usually take single digit microseconds). What NTV does is let them keep dropping the voltage further. At .28v, your 1 watt of leakage power drops down to .12 watts, and you can switch in and out of that state with much less overhead. You can use it to better use brief idle periods where the on/off transition takes too long, and you can use it for cheaper wakeups where the processor only needs to turn on for a brief period for short, low level operations.
|
|
#
¿
Feb 22, 2012 07:12
|
|
|
Factory Factory posted:The GPU can be completely power gated, in which case it's completely powered down when not in use. Not sipping a milliwatt. To be pedantic, power gating transistors (like all transistors) still leak a little power. Not a lot, and there are techniques to minimize it, but it could potentially be more than a milliwatt. Fuzzy Mammal posted:In hindsight I think that dude is just butthurt that nobody cares about linux? Oh yes, very much so.
|
|
#
¿
Mar 9, 2012 21:49
|
|
|
And of course, it only does you any good if you can actually make use of six cores. You'll get the best bang for your buck if you get the cheapest current generation i5, or i7 if you need heavily threaded performance. There are so many better ways to spend the price difference between the mid-line and high-end processors.
|
|
#
¿
Jun 1, 2012 17:25
|
|
|
Shaocaholica posted:According to this, the OS is actually in control of turbo boost: That's not actually what that means. What that really means is turbo boost won't run when the OS has told SpeedStep to underclock the processor. The CPU itself is totally in control of whether or not it overclocks itself.
|
|
#
¿
Jul 5, 2012 20:17
|
|
|
Apple controls their hardware. If Apple wants a feature, they tell their suppliers to make it, and it gets done, whether it's a ratified part of a standard or not (see also: mSATA).
|
|
#
¿
Jul 6, 2012 21:23
|
|
|
Shaocaholica posted:If so, seems odd new tech would be introduced to the server/workstation market first. It does? The server market has a lot of niche needs and demand for high performance, and is willing to pay high dollar for it.
|
|
#
¿
Aug 2, 2012 19:25
|
|
|
Factory Factory posted:In a laptop APU, sure, the lower power consumption, increased bandwidth, and eventually higher density would be great. Even then, moving to a faster external memory interface isn't the answer. You want to solder some LPDDR2 right on top of your GPU. When your bus is only a millimeter long, with no routing issues, you can afford to make it ridiculously wide (512 bit or even 1024 bit) to make up for low frequencies, and you get massive power and latency savings in the process. And you get big cost and space savings when you don't have to put that stuff on the motherboard, to boot.
|
|
#
¿
Aug 4, 2012 20:40
|
|
|
Aside from price (and probably cable durability as well), it's a lot easier to send power over copper.
|
|
#
¿
Sep 4, 2012 15:26
|
|
|
Puddin posted:I have a thermaltake 600w PSU I'm pretty sure everything Thermaltake sells is complete poo poo, so I'd recommend replacing it regardless (With a SeaSonic X650, because everything SeaSonic sells is awesome).
|
|
#
¿
Feb 22, 2013 22:17
|
|
|
Alereon posted:I think we can also expect better clockspeed and voltage scaling due to improvements in how the 22nm FinFET process is utilized thanks to the lessons learned from Ivy Bridge. This cuts both ways, though - Ivy Bridge benefits in yields (and thus headroom) from the lessons learned from Sandy Bridge, while the new architecture has new critical paths that aren't yet optimized (this is the beauty of the Tick/Tock cadence). Chuu posted:That absolutely does not seem like an enterprise feature to me, it sounded more like a free performance boost with a recompile for applications with poorly written threaded code. That sounds exactly like an enterprise feature to me. Consumer applications are extremely slow to uptake new processor features, because it generally isn't easy to support the new feature without breaking compatibility with old processor revisions.
|
|
#
¿
Mar 19, 2013 22:28
|
|
|
|
| # ¿ Apr 24, 2024 11:37 |
|
|
That's an awfully big die for 64MB of RAM on any modern process. I wouldn't be surprised if that's 256MB, even.
|
|
#
¿
Apr 12, 2013 22:58
|
|